An Article from the March 2002 JOM: A Hypertext-Enhanced Article

The authors of this article are with the Department of Materials Science and Engineering at the University of Wisconsin–Madison.
Exploring traditional, innovative, and revolutionary issues in the minerals, metals, and materials fields.





















Overview: Amorphous Materials

Amorphous Aluminum Alloys—Synthesis and Stability

John H. Perepezko and Rainer J. Hebert

Recent innovations in metallic glasses have led to new alloy classes that may be vitrified and a re-examination of the key alloying factors influencing glass formation and stability. The new alloy classes are usually at least ternary systems and often higher order that can be grouped into two general categories. In one case, large, bulk volumes may be slowly cooled to the glassy state, which signifies a nucleation controlled synthesis. The other important class is represented by aluminum- and iron-based glasses that can be synthesized by rapid solidification processes such as melt spinning. These glasses are often called marginal glass formers that are synthesized under growth-controlled kinetic conditions. Glasses in both alloy classes can also be synthesized by intense deformation of crystalline multilayer arrays. These developments represent a major level of microstructure control that has an impact on the structural performance and stability.


Since their discovery in 1960 by Duwez1, amorphous metallic alloys have been considered attractive materials with a number of unique properties. They remained mostly of academic interest in laboratory studies, however, until the commercialization of iron-based glasses for magnetic applications. Recently, two advances have attracted a resurgence of attention in metallic glasses in terms of both fundamental understanding and potential functional and structural applications.

One advance is the discovery that, with certain multicomponent alloys involving zirconium-, iron-, and magnesium-base systems, it is possible to achieve the amorphous state in bulk volumes during slow cooling of the liquid or during intense deformation of layered component assemblies. The second development concerns a growing number of alloys, including aluminum-base and iron-base alloy systems, that can be processed as glasses by rapid liquid quenching. One remarkable characteristic of the latter systems is that the alloys contain >80 at. % of the base component and do not have a deep eutectic, which has been a common guideline for easy glass formation.2 Instead, a multicomponent combination of constituents with large atomic size differences (i.e., >12%)3 and a negative heat of mixing appear to be key factors favoring glass formation. For amorphous aluminum alloys, glass formation is favored for multicomponent compositions covering typical ranges of aluminum (80–92 at.%), rare-earth (RE) addition (3–20 at.%) and transition metal (TM) components (1–15 at.%) as illustrated in Figure 1a.2 Limited glass formation is also found in binary Al-RE alloys such as the Al-Y system shown in Figure 1b between 8–12 at.% yttrium and in aluminum alloys with multiple TM additions.2 Moreover, the initial transformation reaction upon devitrification for aluminum-rich compositions (>88 at.% aluminum) in this class is a primary crystallization of the base component that yields a microstructure consisting of aluminum nanocrystals with diameters ranging from 5 nm to 20 nm, with volume fractions of up to 0.3 dispersed within an amorphous matrix.

This microstructure, which can be classified as a nanophase composite, is clearly responsible for the unprecedented properties of high strength levels of 1,200—1,500 MPa. A useful perspective on the structural performance of aluminum-based systems based upon the relationship between tensile strength and microstructural scale has been presented by Greer4 and was adapted in Figure 2 to provide a comparison to some other common high-strength alloys. It is evident from Figure 2 that, even in terms of a direct comparison of strength values, the nanostructured aluminum alloys offer advantages. Moreover, these advantages are enhanced when the relative density is considered as a basis for a specific strength comparison. In the analysis of the exceptional strength, the high solute content of the nanostructured aluminum alloys compared to crystalline aluminum alloys appears to be one important factor,5 but the interaction between the aluminum nanocrystal dispersion and the shear bands that develop upon deformation of the amorphous matrix is also likely to be important.2

These advances have been established by the pioneering by Inoue at Tohoku University6,7, but there have also been major contributions from Johnson at Caltech8,9 and Shiflet and Poon at the University of Virginia.10,11 Both the development of bulk metallic glasses and rapidly quenched glasses that undergo primary crystallization to nanocrystalline dispersions are exciting advances, but they also raise a number of intriguing fundamental issues. The new classes of metallic glasses are closely related to two important aspects of solidification that involve kinetic competition: avoidance of crystallization upon cooling of the liquid and control of crystallization upon heating of the glass. Although there are connections between these aspects, including the common underlying important role of melt undercooling as a measure of liquid metastability, in each case the controlling reactions occur under regimes of vastly different kinetic constraints.12 Similarly, with solid-state deformation-induced alloying, the stored energy due to defects, grain refinement, and solute supersaturation are crucial measures of the level of metastability in the analysis of amorphization and the development of nanostructured microstructures.13

The focus of this article is on the development of nanostructured materials through controlled primary crystallization reactions of amorphous alloys. Other devitrification reactions that yield nanostructured intermetallic phases and quasicrystalline phases are not discussed, but are also promising in terms of their structural and functional performance. Some of the key issues concerning synthesis and stability are illustrated from the observed behavior in amorphous aluminum alloys, but the discussion also applies to other similar alloy systems, such as iron-base alloys.

Figure 1a
Figure 1b
Figure 2

Figure 1. (a-left) The composition ranges of reported glass formation in Al-Y-M alloys.3 The relative ease of glass formation and the thermal stability of amorphous alloys varies within the glass-forming ranges; (b-right) The Al-rich portion of the Al-Y phase diagram.
Figure 2. The relationship between the microstructural scale and room-temperature tensile strength for Al-base alloys. Strength values for a Ti alloy and an ultra-fine grained steel are shown for comparison (adapted from Greer4).



The kinetic transition from crystalline products to an amorphous phase is a common structural change that occurs during solidification with an increasing cooling rate and liquid undercooling. Often, the transition is represented by a critical cooling rate and interpreted as a sharp structural change.14 However, there are also many reports of mixed crystal/glass phase structures indicating that the transition occurs over a range of cooling rates reflecting the kinetic competition. Indeed, following amorphization by rapid melt quenching, many metallic glasses do not exhibit a clear glass-transition signal, Tg, upon reheating. It is useful to note that the glass transition is not a phase transformation in a thermodynamic sense, but it is a kinetic manifestation of the slowing down of atomic transport in the liquid with cooling. In fact, the calorimetric glass-transition signal is due to the large change in heat capacity that occurs when a liquid becomes configurationally frozen. The slowing down of atomic transport is also reflected by an increase in liquid viscosity. The time for the liquid structure to relax during cooling is related to the viscosity, and, for typical laboratory measurement conditions, Tg corresponds to a viscosity in the range of 1012–1013 poise (1011–1012 Pa-s).

Figure 3a
Figure 3b
Figure 3c
Figure 3d

Figure 3. TEM bright-field images from Al88Y7Fe5 melt-spun ribbons that were isothermally annealed at 245°C for (a) ten minutes; (b) 30 minutes; (c) 100 minutes; and (d) continuous heating DSC trace at 40 K/min showing a
primary crystallization onset at 276°C.

When a glass transition is not observed during heating of a metallic glass, exothermic maxima develop, indicating a multiple-stage crystallization15,16, as shown in Figures 3a, 3b, 3c, and 3d for an amorphous Al88Y7Fe5 ribbon. Microstructural analysis has established that for many aluminum-base classes, the initial crystallization corresponds to primary phase formation (i.e., aluminum).17 This behavior is of importance in understanding the kinetic control of glass formation. The two basic strategies to synthesize amorphous alloys are illustrated schematically in Figure 4. With nucleation control, the undercooling that is achieved during cooling bypasses the nucleation reaction [nucleation rate (N)] and the nucleation size distribution,18 C(n), that may be retained by the cooling does not overlap with the critical nucleus size, n*, at the crystallization temperature, Tx. As a result, there is no precursor reaction to influence the evolution of crystalline clusters during subsequent thermal treatment. In this way, a clear separation in temperature between the Tg and Tx signals can be observed during reheating of a glass. These kinetic conditions are the basis for bulk glass formation during slow cooling. During isothermal annealing at Tx, the heat-evolution rate, Q, exhibits a clear delay before the onset of the nucleation reaction and a peak maximum associated with nucleation and continued growth. On the other hand, under growth-control conditions, because the cooling rate is insufficient to bypass the nucleation onset completely, some small fraction of crystallites may form initially. However, the rapidly rising viscosity and falling growth rate, G, with continued cooling near Tg, prevents rapid cluster growth. In this case, as indicated in Figure 4, upon reheating a sample with pre-existing crystallites (i.e., quenched-in nuclei), rapid crystallization due to the development of quenched-in clusters as well as additional nucleation ensues at Tx. While many of the early metallic glass alloys were synthesized under growth-controlled conditions (i.e., marginal glass-formers)19 the primary crystallization-particle densities in these alloys are of the order of 1018 m–3. For the new class of amorphous aluminum- and iron-base glasses, the primary crystallization-number densities range from 1021 to almost 1023 m–3. Both of the mechanisms for glass formation that are outlined in Figure 4 can yield a high number density of nanocrystals upon devitrification.

Figure 4
Figure 5

Figure 4. The principal forms of kinetic control for metallic glass formation.
Figure 5. A schematic of the liquid viscosity behavior vs. Tg / T for strong and fragile glasses.


Another important distinction between alloys that form bulk glasses and the marginal glass-forming alloys is based upon the temperature dependence of the liquid viscosity.20 The main features of the viscosity behavior are shown in Figure 5, where it is evident that “strong” liquids display an Arrhenius type of temperature dependence. A good example of a strong liquid is SiO2, but the metallic bulk-glass-forming alloys also display strong liquid characteristics.20 For the “fragile” liquid behavior shown in Figure 5, the viscosity is low even in the undercooled regime, but increases sharply upon approaching the glass transition. It appears that the marginal glass-forming alloys exhibit a fragile type of viscosity behavior. It is evident that the transport behavior will impact both the ease of glass formation and the kinetics of nanocrystal development. The different synthesis routes that are shown in Figure 4 originate from the relative time scales for the onset of nucleation and melt cooling. The transition from growth control to nucleation control can be achieved either by an increase in the cooling rate or by lengthening the time for onset of nucleation, tn. Since tn is related to atomic transport in the liquid, it is evident that strong liquids with high viscosity are favored for bulk-glass formation. It is also apparent that tn can be lengthened by removing active nucleation sites from the melt. In fact, this is the basis for the effectiveness of melt-fluxing, which has been shown to promote bulk-glass formation.21,22 The actual mechanism for the development of the ultrahigh number densities of nanocrystals is under active study, and proposals based upon homogeneous19,23 and heterogeneous24,25 nucleation and precursor phase separation reactions26 are under examination.

The attainment of nanocrystal dispersions of essentially pure aluminum with ultrahigh number densities is important, but equally important is the high thermal stability. One indication of this stability is the wide temperature range between the primary crystallization and final crystallization of 75–100°C in Figures 3a, 3b, 3c, and 3d. Within this range, there is a metastable two-phase coexistence involving the aluminum nanocrystals and the surrounding amorphous matrix, with little coarsening of the microstructure. The complete analysis of the stability is under investigation, but it is clear that at least part of the sluggish kinetics is related to the large differences in component atom sizes and diffusivities27–29 as well as the onset of impingement of the diffusion fields from neighboring nanocrystals.17 Indeed, even at a particle density of 1021 m–3, the average nanocrystal separation is only about 100 nm. It is also evident that, in order for the aluminum nanocrystals to grow, there is a rejection of solute (i.e., TM and RE) as is typical for primary crystallization reactions. The low solute diffusivities, especially for the large RE atom, act to limit the growth30. For example, as shown by transmission electron microscopy (TEM) analysis, aluminum nanocrystals are growing at temperatures below the calorimetrically determined crystallization onset, but the growth rate is too slow to yield a measurable heat-evolution rate.31-32 The reaction rate increases with continued heating to the peak onset. At the same time, with the high particle density, the diffusion fields due to the solute rejected during nanocrystal growth impinge soon after the crystallization onset. This kinetic restriction inhibits further nanocrystal growth and accounts for the asymmetric crystallization peak and the remarkable thermal stability.17

Figure 6a

Figure 6b

Figure 6. TEM bright-field images of Al87Y7Fe5Pb1 (a-top) as-spun ribbon showing nanoscale Pb particles within an amorphous matrix; (b-bottom) after isothermal anneal at 245°C for ten minutes showing Pb particles and a high density (3.1 × 1022 m–3) of Al nanocrystals in an amorphous matrix.


The limited experimental information available indicates that the nucleation process during primary crystallization of amorphous aluminum alloys is heterogeneous in nature, but the origin of the active catalytic sites is uncertain.33 In addition, with a number of amorphous iron alloys, the development of a high density of iron nanocrystals is strongly promoted by the addition of small amounts of certain solutes, such as copper, that cluster to catalyze iron nanocrystal nucleation.34-36 In amorphous aluminum-base alloys, a comparable nucleation catalysis behavior can be observed with both soluble and insoluble additions. For example, the addition of 1 at.% copper to amorphous Al88Y4Ni8 has been reported to yield an aluminum nanocrystal density approaching 1023 m–3, with diameters of 5–7 nm.37 Similarly, the incorporation of nanosized lead-crystalline particles in an amorphous matrix is effective in catalyzing the crystallization of aluminum-nanocrystals and yields a significant increase in the total number density of nanocrystals.38 The TEM images in Figures 6a and 6b show that, in as-spun Al87Y7Fe5Pb1 ribbon, the sample is predominantly amorphous with discrete spherical regions of crystalline lead with a particle density of about 1.6 × 1021 m–3. During annealing, each lead particle nucleates an aluminum nanocrystal. Aluminum nanocrystals have also been observed to develop independently in the amorphous matrix. A comparison of the particle-number density achieved in the annealed Al88Y7Fe5 and Al87Y7Fe5Pb1 melt-spun samples (Figures 3a and 6b), demonstrates that the incorporation of lead increases the particle density from 2.7 1021 to 3.1 1022 m–3. The catalytic effect upon nanocrystallization of a-Al opens up a new opportunity in synthesizing bulk amorphous alloys with an ultrahigh number density of nanoscale dispersoids.


One consequence of the metastability of amorphous phases is that the structure and properties of amorphous materials can depend on the processing pathway. This pathway dependence offers the chance to obtain amorphous phases with novel characteristics that cannot be achieved by melt-spinning. Amorphization reactions during mechanical mixing, such as cold-rolling39, can occur after continued folding and rolling of initially crystalline multilayer samples, as illustrated schematically in Figure 7. The process can yield a true strain in the multilayer sample of the order of 100. The results of a cold-rolling experiment with an Al-Sm multilayer sample with a nominal composition of Al92Sm8 are depicted in Figure 8. The crystalline sample transforms to a partially amorphous sample after 80 rolling and folding passes.40 The transformation occurs continuously, without a noticeable exothermic reaction during rolling at a low strain rate (<10–2 s–1). The continuous-heating differential scanning calorimeter (DSC) curve, depicted in Figure 8, reveals a glass-transition at a temperature of 171°C. The continuous-heating curve of a melt-spun Al92Sm8 sample is also shown in Figure 8. For the melt-spun sample, which appears to be fully amorphous according to standard x-ray diffraction and TEM analysis, no glass-transition is observed in the DSC experiment because it is concealed by the primary crystallization of a-Al.40

The comparison between the melt-spun Al92Sm8 ribbon and the cold-rolled Al92Sm8 multilayers demonstrates that, for a given composition, two different amorphous phases can be accessed by the two processing routes. Similar experiments have been performed with other glass-forming alloys, such as a Zr-Ni-Cu-Al multilayer sample with a nominal composition of Zr68Al5Ni9Cu18. After approximately 80 folding and rolling passes, an amorphous structure is obtained.41 Moreover, the continuous heating DSC curve indicates a glass transition at the same temperature as the melt-spun counterpart. These results can be rationalized in terms of the different synthesis routes and kinetics shown in Figure 4. Similar to the Al88Y7Fe5 marginal glass former, the vitrification behavior of the Al92Sm8 sample is growth controlled. The melt-spinning process retains a high density of clusters in the sample. The rapid increase in viscosity upon cooling prevents the growth of the clusters. In contrast, no significant cluster concentration is retained in the zirconium-based glass during melt-spinning.

The rolling experiments indicate that the cold-roll technique also has the potential to yield glass phases for alloy systems and compositions that do not form glasses by melt-spinning.42 In mechanical amorphization processes, the amorphization reactions occur at the multilayer interfaces. A rapid increase in the interface area is, therefore, beneficial for amorphization.

Figure 7

Figure 7. A schematic illustration of the cold-roll and fold process. A multilayer of elemental foils with foil thicknesses between 7 µm and 25 µm is reduced by 50% with each rolling pass.

Figure 8

Figure 8. A comparison of TEM bright-field images and DSC continuous heating (40 K/min) traces for cold-rolled (upper half) and melt-spun (lower half) Al92Sm8 samples. The melt-spun sample was processed at a wheel speed of 55 m/s.

The difference between the amorphous phases in the cold-rolled and the melt-spun Al92Sm8 samples has its origin in the lack of quenched-in nuclei in the rolled amorphous phase. This raises the question whether quenched-in nuclei in marginal glass-formers are stable against rolling or if they disintegrate. Although intuition may suggest that the rolling process would enhance disorder, the opposite effect is observed for marginal glass-formers. The core tendency of the rolling experiments is an initial crystallization of a-Al with continued folding and rolling of the ribbons. The kinetics of this crystallization process appear to be linked to the initial size distribution and density of the quenched-in nuclei in the amorphous matrix. In some cases, for example in melt-spun Al92Sm8, the sample fully crystallizes during rolling. Amorphous aluminum alloys that follow the nucleation-controlled solidification pathway and, therefore, have no significant quenched in cluster concentration, are considerably more stable against a rolling-induced crystallization reaction. The primary crystallization of marginal glass-formers during initial rolling implies not only a transport of the constituents through the amorphous matrix, but also a preferential clustering of aluminum atoms. A key question to be answered is if the redistribution of solute atoms during crystallization is determined by an athermal mechanical process or by a thermally activated process. In the same vein, more attention is needed for the complete structural characterization of the as-spun material. The existence of barely detectable quenched-in nuclei in the as-spun material has an impact on the response to deformation. Understanding the nature of shear-induced crystallization, thus, necessitates a refined knowledge of the amorphous structure prior to deformation.


The recent discoveries of new multicomponent alloy glasses that can be vitrified either by slow cooling of the melt as bulk glasses or by rapid melt quenching as marginal glasses or by intense deformation of multilayers represent a major advance. The developments have led to a reconsideration of the fundamentals of the alloying behavior favoring easy glass formation and a resurgence of effort in discovering new glass-forming alloys.

In addition to the use of glass-forming alloys as single-phase material, they have been demonstrated to serve as most effective precursors for the development of nanostructured materials in bulk form. Indeed, the controlled primary crystallization of amorphous aluminum alloys yields essentially a nanophase composite of nanocrystalline aluminum dispersed within an amorphous matrix. Similar microstructures can be developed in the iron-based and other amorphous alloys. It is remarkable that the nanocrystal number density can achieve high levels of 1021–1023 m-3 to yield ultra-high strength.

An equally remarkable observation is the relative high thermal stability of the nanophase composite. This is truly a novel microstructure that has revealed challenging fundamental issues on the kinetics and deformation processes that control the structure synthesis and performance.


The continued support of the Army Research Office (DAAD 19-01-1-0486) and encouragement from W. Mullins for studies of novel microstructure synthesis during solidification and intense plastic deformation is most gratefully appreciated. The authors are pleased to acknowledge H. Sieber (Universität Erlangen, Germany), G. Wilde (INT, Karlsruhe, Germany), and R. Wu (Intel Corporation) for productive collaboration.


1. W. Klement, R.H. Wilens, and P. Duwez, Nature, 187 (1960), p. 869.
2. A. Inoue, Progr. Mat. Sci., 43 (1998), pp. 365–520.
3. T. Egami and Y. Waseda, J. Non-Cryst. Solids, 64 (1984), pp. 113–134.
4. A.L. Greer, Nature, 368 (1994), pp. 688–689.
5. A.L. Greer, Mat. Sci. Engr., A304-306 (2001), pp. 68–72.
6. A. Inoue, Mat. Trans. JIM, 36 (7) (1995), pp. 866–875.
7. A.P. Tsai et al., Acta Mater., 45 (4) (1997), pp. 1477–1487.
8. W.L. Johnson, Mater. Sci. Forum, 225-227 (1996), pp. 35–49.
9. W.L. Johnson, Mat. Res. Soc. Symp. Proc., 554 (1999), pp. 311–339.
10. Y. He et al., Acta Metall., 41 (2) (1993), pp. 337–343.
11. H. Chen et al., Nature, 367 (1994), pp. 541–543.
12. W.J. Boettinger and J.H. Perepezko, Rapidly Solidified Alloys: Processes, Structures, Properties, Applications, ed. H.H. Liebermann (Parsippany, NJ: Marcel Decker Inc., 1993), p. 17.
13. J.H. Perepezko and G. Wilde, Ber.Bunsenges. Phys. Chem., 102 (1998), pp. 1074–1082.
14. M.T. Clavaguera-Mora, Ber. Bunsenges. Phys. Chem., 102 (1998), p. 1291–1297.
15. L. Battezzati et al., Mat. Sci. and Eng., A179/180 (1994), p. 600–604.
16. Y. He, S.J. Poon, and G.J. Shiflet, Scripta Met., 22 (1988), pp. 1813–1816.
17. D.R. Allen, J.C. Foley, and J.H. Perepezko, Acta Mater., 46 (1998), pp. 431–440.
18. K.F. Kelton, A.L. Greer, and C.V. Thompson, J. Chem. Phys., 79 (1983), pp. 6261–6276.
19. A.L. Greer, Acta Metall., 30 (1982), pp. 171–192.
20. C.A. Angell, Science, 267 (1995), pp. 1924–1935.
21. H.W. Kui and D. Turnbull, Appl. Phys. Lett., 47 (8) (1985), pp. 796–797.
22. A.J. Drehman and A.L. Greer, Acta Metall., 32 (1984), pp. 323–332.
23. S. Omata et al., Phil. Mag. A, 76 (1997), pp. 387–412.
24. R.F. Cochrane, P. Schumacher, and A.L. Greer, Mater. Sci. Engr., A133 (1991), pp. 367–370.
25. R. Wu et al., Mat. Res. Soc. Symp. Proc., 644 (2000), L4.7–L4.12.
26. A.K. Gangopadhyay, T.K. Croat, and K.F. Kelton, Acta Mater., 48 (2000), pp. 4035–4043.
27. K. Hono, A. Inoue, and T. Sakurai, Appl. Phys. Lett., 58 (1991), pp. 2180–2182.
28. K. Hono et al., Acta Metall., 40 (1992), pp. 2137–2147.
29. A.L. Greer, Defect Diff. Forum, 129/130 (1996), pp. 163–180.
30. K. Hono et al., Mater. Trans. JIM, 36 (1995), pp. 909–917.
31. J.C. Foley, D.R. Allen, and J.H. Perepezko, Mater. Sci. Engr., A226-228 (1997), pp. 569–573.
32. J.C. Foley, D.R. Allen, and J.H. Perepezko, Scripta Mat., 35 (1996), pp. 655–660.
33. S.K. Das et al., Mater. Sci. Engr., A304-306 (2001), pp. 159–165.
34. Y. Yoshizawa, S. Oguma, and K. Yamauchi, J. Appl. Phys, 64 (10) (1988), pp. 6044–6046.
35. J.D. Ayers et al., Acta Mater., 46 (1998), pp. 1861–1874
36. T. Ohkubo et al., Scripta Mater., 44 (2001), pp. 971–976.
37. S.J. Hong, P.J. Warren, and B.S. Chun, Mater. Sci. Engr., A304-306 (2001), pp. 362–366.
38. J.H. Perepezko, D.R. Allen, and J.C. Foley, U.S. patent 6,261,386 B1 (University of Wisconsin-Madison, Madison, WI, 17 July 2001).
39. R.S. Mishra et al., editors, Ultrafine Grained Materials (Warrendale, PA: TMS, 2000).
40. G. Wilde, H. Sieber, and J.H. Perepezko, Scripta Mater., 40 (1999), pp. 779-–783.
41. A. Sagel, H. Sieber, and J.H. Perepezko, Acta Mater., 46 (1998), pp. 4233–4241.
42. R.J. Hebert and J.H. Perepezko, Mater. Sci. Forum 386-388 (2002), pp. 21–26.

For more information, contact J.H. Perepezko, University of Wisconsin–Madison, Department of Materials Science and Engineering, 1509 University Avenue, Madison, Wisconsin 53706; (608) 263-1678; fax (608) 262-8353; e-mail

Copyright held by The Minerals, Metals & Materials Society, 2002

Direct questions about this or any other JOM page to

If you would like to comment on the March 2002 issue of JOM, simply complete the JOM on-line critique form
Search TMS Document Center Subscriptions Other Hypertext Articles JOM TMS OnLine